3778±3787
Nucleic Acids Research, 2002, Vol. 30 No. 17
ã 2002 Oxford University Press
Structural insights by molecular dynamics
simulations into differential repair ef®ciency for
ethano-A versus etheno-A adducts by the human
alkylpurine-DNA N-glycosylase
Anton B. Guliaev, Bo Hang and B. Singer*
Donner Laboratory, Life Sciences Division, Lawrence Berkeley National Laboratory, University of California,
Berkeley, CA 94720, USA
Received May 3, 2002; Revised and Accepted July 8, 2002
ABSTRACT
1,N 6-ethenoadenine adducts (eA) are formed by
known environmental carcinogens and found to be
removed by human alkylpurine-DNA N-glycosylase
(APNG). 1,N 6-ethanoadenine (EA) adducts differ
from eA by change of a double bond to a single
bond in the 5-member exocyclic ring and are formed
by chloroethyl nitrosoureas, which are used in cancer therapy. In this work, using puri®ed recombinant
human APNG, we show that EA is a substrate for
the enzyme. However, the excision ef®ciency of EA
was 65-fold lower than that of eA. Molecular dynamics simulation produced similar structural motifs for
eA and EA when incorporated into a DNA duplex,
suggesting that there are no speci®c conformational
features in the DNA duplex which can account for
the differences in repair ef®ciency. However, when
EA was modeled into the APNG active site, based
on the APNG/eA-DNA crystallographic coordinates,
in structures produced by 2 ns molecular dynamics
simulation, we observed weakening in the stacking
interaction between EA and aromatic side chains of
the key amino acids in the active site. In contrast,
the planar eA is better stacked at the enzyme active
site. We propose that the observed destabilization
of the EA adduct at the active site, such as reduced
stacking interactions, could account for the biochemically observed weaker recognition of EA by
APNG as compared to eA.
INTRODUCTION
The saturated exocyclic adduct of adenine, 1,N6-ethanoadenine (EA) (Fig. 1A), has been identi®ed as one of the
products of the reaction of 1,3-bis(2-chloroethyl)nitrosourea
(BCNU) with DNA (1,2). BCNU belongs to the family of
therapeutic nitrosourea compounds used in cancer treatment.
The ethano adducts in DNA structurally resemble the
exocyclic etheno adducts formed from the reaction of the
chemical carcinogen vinyl chloride with DNA (3,4) or by lipid
peroxidation (5). The etheno adducts, particularly 1,N6ethenoadenine (eA), have been extensively studied biochemically and structurally (6). It has been shown that this
adduct can be ef®ciently removed from DNA by rodent or
human alkylpurine-DNA N-glycosylase (APNG) (also termed
alkyladenine DNA glycosylase, AAG) (7±10). The mechanism of eA excision by APNG has been proposed based on the
Ê crystal structure of an APNG mutant protein (E125Q)
2.1 A
complexed to eA-containing DNA (11). Crystallization of the
protein±substrate complex was made possible by substitution
of Glu125 with a glutamine residue, which prevents activation
of the active site bound water acting as a nucleophile. The
authors showed that ¯ipped-out eA has the ability to stack in a
stable position between the aromatic side chains in the enzyme
active site (11). The position of the adduct was also stabilized
by a key hydrogen bond between the main chain of His136 and
N9 of eA, which offered a unique acceptor lone pair essential
for hydrolysis of the C1¢±N glycosylic bond. The His136 side
chain forms hydrogen bond interactions to Tyr157 and the
phosphate group of eA (Fig. 1B).
The ethano adducts differ from etheno adducts by the
change of a double bond to a single bond in the 5-member
exocyclic ring (Fig. 1A). In this work we have addressed the
issue of whether such a small structural change could affect
the recognition and repair ef®ciency of EA compared to eA by
human APNG. Recent work in this laboratory showed that a
small structural change in the adduct structure has an effect on
DNA glycosylase activity (12). Addition of a hydroxymethyl
group to the C8 position of 3,N4-ethenocytosine (eC) to form
8-(hydroxymethyl)-3,N4-ethenocytosine (8-HM-eC), a product of the reaction with the mutagen/carcinogen glycidaldehyde, reduced the repair ef®ciency by Escherichia coli
mismatch uracil-DNA glycosylase (Mug) by 2.5-fold as
compared with that of the structurally related eC. However,
molecular dynamics simulation showed similar alignment and
hydrogen bonding patterns for both adduct pairs in the 25mer
oligomer duplexes used in the biochemical studies (12). The
lower Mug activity toward 8-HM-eC suggests some degree of
steric hindrance to the binding or catalytic activity as a result
of the hydroxymethyl group on the etheno ring.
*To whom correspondence should be addressed. Tel: +1 510 642 0637; Fax: +1 510 486 6488; Email: [email protected]
Nucleic Acids Research, 2002, Vol. 30 No. 17
3779
MATERIALS AND METHODS
Oligonucleotides
Synthesis of the EdA phosphoramidite and its site-speci®c
incorporation into oligonucleotides was described by
Maruenda et al. (13). The edA phosphoramidite was purchased from Glen Research (Sterling, VA). Both derivatives
were placed in the sixth position from the 5¢-end of a 25mer
sequence (X): 5¢-CCG CTX GCG GGT ACC GAG CTC GAA
T-3¢. The unmodi®ed 25mer and complementary strands with
T opposite the modi®ed base were purchased from Operon
Technologies (Alameda, CA). All the oligomers were puri®ed
by HPLC and denaturing PAGE.
DNA glycosylase assay
Figure 1. (A) Chemical structures of the eA and EA adducts. (B) The
APNG active site structure showing the stacking between eA and aromatic
side chains of Tyr127, His136 and Tyr157. The black dashed lines show
hydrogen bonds between eA N9 and His136 N, Tyr157 O4 and His136 Np
and eA OP1 and His136 Nt. The picture was generated using the atomic
coordinates of the crystallized eA-DNA/APNG complex [PDB ID code 1f4r
(11)].
In this work the repairability by human APNG of EA,
incorporated into a 25mer DNA duplex, was investigated and
compared to the repair ef®ciency of eA by the same enzyme
using a DNA glycosylase assay. EA was found to be a
substrate for the human enzyme, but a much weaker one than
eA. The observed difference in rate of excision of eA versus
EA adducts was correlated with the structural data obtained by
molecular modeling. The availability of crystal data for the
APNG enzyme complexed to eA-containing DNA allowed us
to use it as a starting point in our molecular modeling. The
observation of structural perturbations caused by replacement
of eA by EA in the enzyme active site might have an effect on
the substrate preference of eA over EA. The complementarity
between the substrate and enzyme active site should be one of
the factors responsible for the catalytic speci®city and
ef®ciency of repair. However, a number of other events,
such as initial lesion binding/recognition, ease of rotating the
damaged base from the DNA ladder and stabilization of the
extrahelical conformation, also contribute to the ef®ciency of
repair for a particular adduct. These factors can be in¯uenced
by the conformational features of the adduct-containing
duplexes. To evaluate the effects of the EA adduct on
the local and global structural features of the DNA duplex
we performed simulation of an EA-T-containing 25mer
DNA duplex. These data were compared to the eA-T- and
A-T-containing duplexes.
The enzymatic assay used to test APNG-mediated cleavage of
EA or eA from oligonucleotides was carried out essentially as
previously described (14,15). Brie¯y, 25mer oligonucleotides
were 5¢-end-labeled with [g-32P]ATP (speci®c activity
6000 Ci/mmol, 1 Ci = 37 GBq; Amersham Pharmacia
Biotech) and annealed to a complementary strand in a 1:1.5
molar ratio. The standard reactions (10 ml) contained 2 nM
5¢-32P-end-labeled oligomer duplex in 10 mM HEPES±KOH,
pH 7.4, 100 mM KCl, 1 mM EDTA, 1 mM EGTA, 1 mM
DTT, 0.1 mg/ml acetylated BSA and varying amounts of
human APNG protein (a gift from Dr Tim O'Connor,
Beckman Research Institute, Duarte, CA) (16). In these
reactions, a 5¢ AP endonuclease, the major human AP
endonuclease (HAP1) (a gift from Dr Ian Hickson, Oxford
University, Oxford, UK), was added to cleave the apurinic
(AP) site resulting from the excision of EA or eA by APNG
protein. The reactions were stopped by adding equal amounts
of F/E solution (90% formamide plus 50 mM EDTA) and then
heated at 95±100°C for 3 min. Reactions were then resolved
by 12% polyacrylamide±8 M urea denaturing PAGE. For band
quantitation, a Bio-Rad FX molecular phosphorimager and
Quantity One software (v.4.0.1) were used.
Molecular modeling
EA- and eA-containing 25mer DNA duplexes. A set of force
®eld parameters for eA was previously developed using an
ab initio quantum mechanical calculation and procedure
described in earlier publications from this laboratory (12,17).
The EA adduct was built by saturating the C7=C8 double bond
in the imidazole ring of eA. Atom-centered charges were
calculated with the RESP module of AMBER 6.0 using the
partial charges obtained by Hartree±Fock calculation using the
6-311G* basis set in the Spartan 5.0 suite (Wavefunction, Inc.,
Irvine, CA). Prior to the charge calculations, the conformation
of EA was geometry optimized using the 6-31G* basis set.
The eA and EA adducts were incorporated in the sixth position
into the 25mer sequence used in biochemical studies (for
sequence see Materials and Methods). The topology and
coordinate ®les for the three DNA duplexes (eA-T-DNA, EAT-DNA and A-T-DNA, used as a control) were generated with
the xLeap module of AMBER 6.0 (18). Forty-eight Na+ ions
were placed around the phosphate groups to neutralize
negative charges, and an aqueous environment was represented by a rectangular water box, which provided no less than
Ê of TIP3P water molecules around the solute. Two
10 A
3780
Nucleic Acids Research, 2002, Vol. 30 No. 17
Figure 2. (Left) Protein-dependent cleavage of a 25mer oligonucleotide containing either EA or eA by human APNG protein. Increasing amounts of APNG
protein (0.3±4.8 ng for EA and 0.01±0.15 ng for eA) were incubated with 2 nM 32P-end-labeled oligomer substrates for 30 min at 37°C. The AP site produced
by DNA glycosylase action was further cleaved by adding HAP1 (5 ng), a 5¢ AP endonuclease, to the reaction mixture. Note that the use of HAP1 alone had
no detectable effect on either EA- or eA-containing templates. (Right) Time-dependent cleavage of a 25mer oligonucleotide containing either EA or eA.
Oligomer duplexes were reacted with 3 (for EA) or 0.15 ng (for eA) APNG protein for varying times at 37°C. The scanning data were normalized as nM
oligomer substrate cleaved per ng APNG protein. (Inset) Detailed time-dependent response of EA excision by APNG.
nanosecond molecular dynamics simulation runs at 310 K,
using particle-mesh Ewald (PME) to treat Coulombic interactions and a 2 fs time step, were generated after the system
achieved the correct density and volume (17).
EA-DNA/APNG and eA-DNA/APNG complexes. In this work
the high resolution X-ray crystal structure of the eA-DNA/
APNG complex (PDB code 1f4r) served as the starting
structure. Hydrogens were added using the xLeap module of
AMBER 6.0. To generate the EA-DNA/APNG complex, eA
was replaced by the geometry optimized EA adduct, using
Insight II (Biosym/MSI, San Diego, CA). Two sets of
topology and coordinate ®les for the APNG protein complexed to eA-DNA and EA-DNA were generated using the
xLeap module of AMBER 6.0. A rectangular box of TIP3P
Ê of explicit
water molecules was added, providing at least 10 A
solvent around each DNA/protein complex, yielding 9292
water molecules. The complete system consisted of approximately 31 728 atoms and has the initial dimensions 73.439,
Ê in the x, y and z directions, respectively.
70.948 and 76.668 A
The initial density of the water around the protein was
0.806 g/cm3. Molecular dynamics simulations were carried
out using the SANDER module of AMBER 6.0 with a 2 fs
time step. SHAKE was applied to all hydrogen atoms and a
Ê cut-off was used for Lennard±Jones interactions.
10 A
Constant pressure was maintained with isotropic scaling. All
long-range electrostatic interactions were handled using the
PME method. In the beginning of the simulations, the water
box was subjected to a series of equilibration molecular
dynamics runs while holding the DNA/APNG complex ®xed,
and was similar to the procedure used for the DNA duplexes.
The equilibration runs began with 1000 steps of minimization
followed by 10 ps of simulation, during which the temperature
was slowly raised from 0 to 310 K and kept at this temperature
for another 50 ps. During the ®rst 30 ps of simulation the water
density and pressure converge to the correct values (1.01 g/cm3
and 1 atm, respectively). This was followed by a second set of
1000 steps of minimization and 3 ps of simulation, which were
carried out with the restraints on the solute molecule reduced
to 25 kcal/mol. Finally, ®ve rounds of 800 steps of conjugate
gradient minimization were performed with the positional
restraints reduced by 5.0 kcal/mol in each round. The
unrestrained molecular dynamics production runs of 2 ns
were initiated after the last round of minimization. The ®nal
structures representing the conformational family for the
DNA/enzyme complexes produced by molecular dynamics
simulation were generated by averaging the molecular
dynamics trajectories based on root mean square deviation
(RMSD) pro®les (from 0.4 to 2 ns).
Structural analysis and calculations. The molecular dynamics
trajectories were processed using the analytical modules of
AMBER 6.0 and visually analyzed with the VMD program
(19). Nucleic acid structural parameters were derived using
CURVES 5.1 (20). Production runs for the 25mer DNA
duplexes and DNA/APNG complexes were carried out on 64
processors (16 processors per node) using the IBM SP RS/
6000 supercomputer available at the National Energy
Research Scienti®c Computing Center, Lawrence Berkeley
National Laboratory. The equilibration runs and trajectory
analysis were performed on a Silicon Graphics Origin 200
server interfaced with a dual processor Octane workstation.
RESULTS
Biochemical assay
We ®rst tested the excision activity of APNG protein towards
EA since this enzyme excises the closely related adduct eA as
well as another ethano adduct, N2,3-ethanoguanine (8,10,21).
As shown in Figure 2 (left), APNG protein showed a proteindependent cleavage of a 32P-end-labeled EA-containing
25mer oligomer duplex (EA-T). The cleavage products
Nucleic Acids Research, 2002, Vol. 30 No. 17
3781
Figure 3. Autoradiogram of gel electrophoresis of 5¢-32P-labeled oligonucleotides after reaction with human APNG for varying times (0±60 min).
The amount of APNG used for these reactions was 3 ng for EA and 0.15 ng
for eA excision. For reaction and gel procedure details see Materials and
Methods. The arrows show the position of the 5mer cleavage product. M, a
5mer marker with the same sequence as the expected cleavage product.
from EA- and eA-containing oligonucleotides after 5¢ AP
endonuclease treatment were both 32P-labeled 5mers are
shown in Figure 3 (arrows). These are the expected products
resulting from the 5¢ hydrolysis by HAP1 of an AP site at the
sixth position. However, the extent of EA excision by APNG
was much lower than that of eA excision, as indicated in
Figure 2 (left).
Excision of EA from the 25mer DNA duplex by APNG
protein as a function of time is shown in Figure 3, in which a
comparison was made between the rate of excision of EA and
eA under the same assay conditions except that the amount of
APNG used was different (3 and 0.15 ng for EA and eA,
respectively). In Figure 2 (right) the scanning results were
normalized per ng protein in order to compare the rates of
these two activities. It is evident that the excision of EA occurs
much more slowly than that of eA, with an ~65-fold
difference.
Conformation of the EA- versus eA-containing duplexes
Two nanosecond molecular dynamics calculations were
performed for the two 25mer DNA duplexes used in the
biochemical studies (eA-T-DNA and EA-T-DNA) and a
corresponding control A-T-DNA (25mer DNA duplex with an
unmodi®ed A at the sixth position). The conformational
stability was evaluated by calculating RMSD values of each
picosecond relative to the coordinates of the initial energy
minimized structures for all three DNA duplexes. RMSD
values for all atoms, the ®ve central base pairs and the adductcontaining base pair are shown in Figure 4A and B for eA-Tand EA-T-DNA, respectively. Based on RMSD values, both
structures reached conformational equilibrium after the ®rst
400 ps and showed a plateau for the rest of the simulation. To
Figure 4. Time dependence of RMS deviations of the 25mer DNA duplexes
containing eA-T (A) and EA-T lesions (B). The data is shown for all
atoms (black), the lesion (red) and 5 bp with the lesion in the middle
(blue) (C4T5eA/EA6G7C8/G47A46T45C44G43). Both structures reached
conformational equilibrium after the ®rst 400 ps.
monitor integrity of the duplex during the simulation we
calculated Watson±Crick hydrogen bond distances and percentage occupancy for all base pairs in the duplexes. All
hydrogen bonds, including 5¢-TA and 3¢-GC base pairs
¯anking the adduct site, were 98±100% occupied during the
entire simulation (data not shown). Terminal bases were not
included in hydrogen bond calculations due to known fraying
effects, which were also observed in our simulations. No
hydrogen bonding was observed in the eA-T and EA-T
mismatches. Top and side views for the T5X6G7/A46T45C44
motifs, where X = A, eA or EA, are shown in Figure 5. In both
lesion-containing duplexes, compared to the unmodi®ed
duplex, the adduct was displaced towards the major groove,
while the opposite T remained stacked between A and C bases.
Figure 6 shows average values for the inter- and intra-base pair
parameters (Fig. 6A and B, respectively) of the 5 bp for the
eA- and EA-containing duplexes and corresponding control.
Average values were calculated over the simulation trajectory.
Presence of the adduct had a similar effect on the conformation of the mismatch and neighboring bases in eA-T-DNA
and EA-T-DNA. A positive shear (SHR) value was observed
for both eA-T and EA-T base pairs, indicating the magnitude
of displacement of the adduct towards the major groove
(Fig. 5). Another two intra-base pair parameters affected by
the presence of either eA or EA in the DNA duplex were
buckle and propeller twist. Considerable propeller twist
(15±23°), compared with the unmodi®ed DNA (<8°) was
observed for the T5-A46, eA/EA-T45 and C8-G43 base pairs in
3782
Nucleic Acids Research, 2002, Vol. 30 No. 17
Figure 5. Top (left) and major groove (right) views for the 3 bp motifs for the A-T-, eA-T- and EA-T-containing 25mer DNA duplexes produced by 2 ns
molecular dynamics simulations. EA and eA adducts are shown in red and A and T bases are shown in blue. Both eA-T and EA-T base pairs produced similar
structural motifs with the adduct displaced toward the major groove and a non-planner alignment of the bases, as compared to the A-T pair. No hydrogen
bonds were observed between the bases in the eA-T and EA-T pairs. Yellow dashed lines show Watson±Crick hydrogen bonds for the A-T base pair.
the lesion-containing duplexes (Fig. 6A). Buckling around the
lesion site was also larger in magnitude than for the
unmodi®ed DNA. Perturbations in inter-base pair parameters,
which probably best describe stacking interactions, were
similar in both the eA and EA duplexes. The most noticeable
differences from the unmodi®ed duplex were observed for the
tilt (TLT), roll (ROL) and twist (TWS) parameters (Fig. 6B).
The high magnitude of TWS for the T5-A46/eA6-T45 and T5A46/EA6-T45 base pair steps (58° and 50°, respectively)
indicated a larger helical twist at these steps compared with
the rest of the adduct-containing and unmodi®ed duplexes.
The succeeding steps, eA6-T45/G7-C44 and EA6-T45/G7-C44,
showed much smaller TWS values (2° and 4°, respectively),
characteristic of untwisting of the DNA at the lesion site. In
previous modeling work from our laboratory (17) we reported
smaller TWS values at the eA-T base pair in a 15mer DNA
duplex. Moreover, the magnitude of TWS was sequencedependent (17). The curvature of the DNA was calculated
using the CURVES 5.1 algorithm and was not affected by
presence of the adduct. To avoid a contribution from the
highly ¯exible DNA ends, the terminal base pairs were not
included in the curvature measurements. The values for the
EA-T-DNA and eA-T-DNA duplexes were 14° and 11°,
respectively. The sugar conformation of EA falls in the C2¢endo conformation, while eA was closer to the C1¢-exo range.
Both adducts stack in an anti orientation into the DNA helix.
The conformation of the eA-T base pair produced by our
modeling approach was similar to the conformation of that
lesion reported based on NMR data (22,23). However, in our
modeling we observed a slightly bigger shift of eA towards the
major groove than was reported by NMR. The displacement of
Ê , while
eA along the x-axis towards the major groove was 2.5 A
Ê , compared to unmodi®ed A in the
EA was displaced by 2.0 A
control duplex. The differences in the conformation of the
eA-T pair between NMR and modeling can be attributed to
sequence-dependent effects (17).
Nucleic Acids Research, 2002, Vol. 30 No. 17
3783
Figure 7. Time dependence of RMS deviations (RMSD) of the eA-DNA/
APNG (A) and EA-DNA/APNG complexes (B). Black, complex (DNA +
enzyme); blue, enzyme alone; green, DNA alone; red, active site. The conformational families produced by molecular dynamics simulation for the
eA-DNA/APNG and EA-DNA/APNG complexes deviate minimally from
the crystal coordinates. High RMSD ¯uctuations for the DNA duplex (green
traces) can be explained by the contribution of the more ¯exible ends.
(C) The RMSD values for the eA (blue) and EA (gray) adducts.
Figure 6. (A) Average values for the intra-base pair parameters describing
the geometry of base pairing for the 5 bp in the A-T-, eA-T- and EA-T-containing duplexes. (B) Average values for the inter-base pair parameters
describing the stacking interactions for the 4 bp steps in the A-T-, eA-Tand EA-T-containing duplexes. The tick marks on the x-axis indicate the
base pair step. For example, label C4-T5 corresponds to the C4-G47/T5-A46
base pair step.
Effect of the EA adduct on the APNG active site
The availability of crystal data for human APNG complexed
to eA-containing DNA allowed us to use this structure as a
starting point in our molecular modeling study in which we
addressed the question of substrate preference of this enzyme
for eA over EA. Simple superimposition of EA over eA did
not reveal any conformational effects which EA might have on
the active site of APNG enzyme. First, to validate our
modeling protocol, we performed 2 ns simulation of the
APNG/eA-DNA complex (PDB ID code 1f4r). The analysis of
the overall structure and position of the adduct in the active
site showed that the averaged minimized structure produced
by molecular dynamics simulation deviates minimally from
the crystal coordinates. All averaged RMSD values where
Ê , with a value of 0.9 6 0.06 A
Ê for the enzyme active site,
<2.0 A
Ê for the enzyme, 1.87 6 0.31 A
Ê for the DNA and
1.7 6 0.17 A
Ê for the all-atom RMSD for the entire structure
1.97 6 0.25 A
(Fig. 7A). The largest RMSD ¯uctuations observed for the
DNA duplex bound to enzyme can be explained by the
3784
Nucleic Acids Research, 2002, Vol. 30 No. 17
Figure 8. Superimposition of the eA-DNA/APNG active site from the crystal structure (green) (PDB code 1b4r) over the eA-DNA/APNG active site produced
Ê . The yellow dashed lines indicate the key hydrogen
by 2 ns molecular dynamics simulation (red). The RMSD between active site conformations is <0.91 A
bond between the eA adduct and main chain amide of His136 and two hydrogen bonds which stabilize the position of the His136 side chain. All three
hydrogen bonds remained intact during molecular dynamics simulation.
contribution of more ¯exible DNA ends. All stacking and key
hydrogen bond interactions in the active site remained intact
during this simulation. The superimposition of the active sites
of the crystal structure of eA-DNA/APNG and the eA-DNA/
APNG complex produced by molecular dynamics simulation
is shown in Figure 8. Note that molecular dynamics simulation
produced more pronounced plane-to-plane stacking between
His136 and the imidazole ring of eA than in the crystal
structure.
The RMSD values for the EA-DNA/enzyme complex
showed a similar pro®le to that observed for the eA-DNA/
enzyme complex and indicated overall conformational stability for the system when eA was replaced by the EA adduct
(Fig. 7B). Slightly higher averaged RMSD values were
observed for the active site of the enzyme complexed with
EA-DNA than the active site of the eA-DNA complex (1.1 6
Ê , respectively). Pertinent observations
0.1 versus 0.9 6 0.06 A
can be drawn from monitoring the RMSD values for the
adduct itself during the course of the simulations. Only
corresponding atoms between the structures were compared.
The EA adduct showed signi®cantly higher ¯exibility and
larger deviation from the starting position in our simulation
than the eA adduct (Fig. 7C). The average RMSD value for
Ê , while for the eA adduct it was 0.2 6
EA was 0.36 6 0.13 A
Ê . The main conformational feature observed for the
0.13 A
APNG active site complexed with EA-DNA was displacement
and almost 70° rotation of the His136 side chain. This created
an edge-to-edge packing interaction with EA, rather than the
much more stable face-to-face stacking observed between the
planer eA and His136 in the crystal structure (Fig. 9). Face-toface stacking was also supported during our molecular
dynamics simulation of the eA-DNA/APNG complex. The
change in the stacking interaction between EA and His136
resulted in a weakening of two hydrogen bonds: between the
side chain of His136 and the 5¢-phosphate of EA (eA
OP1±His136 Nt) and the side chain of His136 and
Tyr157 (Tyr157 O4±His136 Np). The evolution of these
hydrogen bonds over simulation time is shown in Figure 10.
However, the key hydrogen bond between N9 of EA and NH
of His136 remained intact in the EA-DNA/APNG complex
(Fig. 10).
DISCUSSION
One of the most important steps in DNA base excision repair
(BER) is recognition and excision of the damaged base from
the DNA ladder by DNA glycosylases. This step is the key
determinant of BER activity against a speci®c lesion. Recent
crystallographic studies revealed details of the base excision
mechanism of DNA glycosylases, including human APNG,
showing that damaged DNA bases are excised by hydrolysis
of the C1¢±N glycosylic bond. The result of this reaction is a
free DNA base and an abasic sugar residue, which is
hydrolyzed by an AP endonuclease, followed by DNA
synthesis and ligation, which restores the correct DNA
sequence (24,25). The position of the adduct in the enzyme
Nucleic Acids Research, 2002, Vol. 30 No. 17
3785
Figure 9. APNG active site structure for the eA-DNA/APNG (blue) and EA-DNA/APNG (yellow) complexes produced by molecular dynamics simulations.
The green dashed lines indicate hydrogen bonds between eA N9 and His136 NH in the APNG/eA-DNA complex and EA N9 and His136 NH in the APNG/
EA complex. The steric clash between the EA exocyclic ring and His136 side chain (indicated by red arrow) resulted in reduced stacking interactions
(edge-to-edge packing between EA and His136) and destabilized the position of the ethano adduct in the enzyme active site.
binding pocket is achieved by rotation of the damaged base out
of the DNA in order that it can be inserted into the enzyme
active site. Correct alignment of the modi®ed base in the
enzyme active site is one of the key steps for successful
removal of that base from the DNA. Structural data on the
enzyme complexed to adduct-containing DNA provide
essential information on the interaction between the substrate
and enzyme active site. The use of molecular modeling has
allowed re®nement of the conformation of DNA/enzyme
complexes with adduct structures, which were not used in the
X-ray crystallography studies. Additionally, structural data on
adduct-containing DNA duplexes should provide valuable
information on some initial steps of BER. Pronounced
structural perturbation around the lesion might be a signal
for the DNA repair enzyme to act on the substrate to prevent
binding to the adduct-containing DNA motif. The stacking
interaction between the adduct and the ¯anking bases, together
with hydrogen bonding with the opposite base, should
in¯uence the ability of the modi®ed base to be ¯ipped out
from the DNA duplex into the enzyme binding site. The
observed conformational features of the adduct-containing
duplexes and enzyme active site bound to the modi®ed base
should be carefully examined and compared with the
biochemical data, thus providing a possible explanation for
differential repair by the particular enzyme.
In this work we have demonstrated that human APNG
recognizes and excises an EA adduct in a de®ned oligonucleotide (Figs 2 and 3). Previously this enzyme was also found to
act on the eA adduct (7,8,10), a structural analog of EA,
although these two adducts are produced by completely
unrelated compounds. Human APNG, as well as homologs in
cells from eukaryotic and prokaryotic species, represents a
family of enzymes with a wide substrate range (for a review
see 26). This work showed that the substrate range of APNG is
still expanding.
Human APNG excises eA from DNA with high ef®ciency
(8,9). We previously reported (9) that eA is even preferred by
Figure 10. Evolution over time of the three hydrogen bond distances in the
APNG binding pocket for the eA- (green) and EA-containing (blue)
DNA/APNG complexes. The eA N9±His136 NH and EA N9±His136 NH
hydrogen bonds remained intact during the entire simulation.
APNG over 3-methyladenine, after which the enzyme was
originally named. The kinetic comparison made in this work
between eA and EA showed that eA is excised much faster
3786
Nucleic Acids Research, 2002, Vol. 30 No. 17
than EA (Fig. 2). Such biochemical data prompted us to
explore the structural basis for the observed difference.
In this work we employed molecular dynamics simulation
to provide structural insights on the EA- and eA-containing
25mer DNA duplexes and the effect of EA on the APNG
active site conformation. Molecular modeling did not reveal
any signi®cant conformational features which can distinguish
between the eA and EA adducts when incorporated opposite
T in 25mer DNA duplexes. Both duplexes have similar
structural motifs around the lesion sites. Both adducts adopted
the anti orientation, were displaced towards the major groove
and formed a non-planar, sheared base pair with the opposite
T. It has been proposed that sheared base pairs can be a
structural feature important for recognition by some DNA
glycosylases (27). No hydrogen bonds were observed between
the bases in the eA-T and EA-T pairs. The sugar pucker of the
EA and eA adducts falls in the C2¢-endo/C1¢-exo range. The
smaller twist values observed for both lesions should
contribute to unwinding of the DNA upon binding to the
repair enzyme. The unwound DNA around the lesion site
allows easy access for the repair enzyme to continue further
adduct recognition and discrimination (28). The overall
conformation of the eA-T base pair produced by modeling
was in general agreement with NMR data on an eA-Tcontaining 9mer duplex (23).
Saturation of the imidazole ring in the EA adduct partially
reduced the stacking ability of this molecule, as compared to
eA, which favors p±p stacking interactions with amino acids
in the enzyme active site. The extra, non-planer hydrogens at
the C7 and C8 positions of EA, as compared to eA, contribute
an additional van der Waals surface area that makes it more
dif®cult to accommodate the adduct between the conformationally constrained Tyr127 and more ¯exible His136. The
replacement of eA by EA in the APNG active site resulted in
an edge-to-edge packing interaction between His136 and EA.
The conformation produced by molecular modeling shows
that in order to accommodate EA in the enzyme active site the
active site required a structural rearrangement involving
His136. A comparison of the APNG crystal structure with
eA-DNA/APNG and abasic pyr-DNA/APNG complexes
showed that the Tyr127, Tyr157 and His136 side chains are
in the same orientation, suggesting that the conformation of
the APNG active site is predetermined and not in¯uenced by
adduct binding (29). The extra energy required to overcome
the steric clash between the aromatic side chain of His136 and
the 7,8-dihydro-imidazole ring of EA should prevent an easy
®t of EA adducts into the APNG active site. However, the
mechanism of EA excision by APNG may be similar to that
reported for eA (11,29). Both eA and EA have a lone pair
acceptor nitrogen (N9), which is unique to the alkylated base.
The position of Glu125 is not changed in the presence of EA
and this residue should be able to deprotonate the active site
bound water for nucleophilic attack on the C1¢ sugar carbon of
EA. The hydroxide nucleophile will be stabilized by Arg182,
the position of which also remains unchanged in the EA-DNA/
APNG complex as compared to the eA-DNA/APNG complex.
Based on the conformations of the eA- and EA-containing
duplexes, it can be suggested that the glycosylase does not
distinguish between these adducts based on local DNA
distortion. Similar structural motifs for these adducts serve
as an initial signal for the enzyme to test the base by forcing
Tyr162 into the helix and displacing the modi®ed base into the
enzyme active pocket. The enzyme active pocket requires
tight interaction between the adduct and the neighboring
amino acids and thus is sensitive to the adduct structure and
conformation. It was shown that an APNG mutant (H136Q),
engineered to eliminate aromatic stacking interactions with
eA, has very low repair ef®ciency (11,29). Moreover, it has
been proposed that base stacking interactions between the
damaged bases and the aromatic side chains of amino acids in
the active site may provide a basis for recognition and excision
by E.coli 3-methyladenine DNA glycosylase II (30,31),
which also excises EA (B.Hang, A.B.Guliaev and B.Singer,
manuscript in preparation). The observed destabilization of
the EA adduct in the active site, such as the weaker stacking
interaction of the adduct with the aromatic side chains of
His136, is likely to contribute to the lower ef®ciency of repair
and explain why this adduct is a less preferable substrate than
eA for human APNG.
ACKNOWLEDGEMENTS
We would like to thank the staff of the National Energy
Research Scienti®c Computing Center (NERSC), Lawrence
Berkeley National Laboratory, especially David Skinner, for
help in setting up AMBER 6.0 on the IBM SP RS/6000
supercomputer at NERSC for our calculations. This work was
supported by NIH grants CA 47723 (to B.S.) and CA 72079
(to B.H.) and was administered by the Lawrence Berkeley
National Laboratory under Department of Energy contract
DE-AC03-76SF00098.
REFERENCES
1. Ludlum,D.B. (1986) Chapter 2. In Singer,B. and Bartsch,H. (eds),
The Role of Cyclic Nucleic Acid Adducts in Carcinogenesis and
Mutagenesis. IARC Scienti®c Publications no. 70. IARC, Lyon, France,
pp. 137±146.
2. Ludlum,D.B. (1990) DNA alkylation by the haloethylnitrosoureas: nature
of modi®cations produced and their enzymatic repair or removal. Mutat.
Res., 223, 117±126.
3. Bartsch,H. (1999) Keynote address. In Singer,B. and Bartsch,H. (eds),
Exocyclic DNA Adducts in Carcinogenesis and Mutagenesis. IARC
Scienti®c Publications no. 150. IARC, Lyon, France, pp. 1±16.
4. Nair,J., Barbin,A., Velic,I. and Bartsch,H. (1999) Etheno DNA-base
adducts from endogenous reactive species. Mutat. Res., 424, 59±69.
5. El Ghissassi,F., Barbin,A., Nair,J. and Bartsch,H. (1995) Formation of
1,N6-ethenoadenine and 3,N4-ethenocytosine by lipid peroxidation
products and nucleic acid bases. Chem. Res. Toxicol., 8, 278±283.
6. Singer,B. and Bartsch,H. (eds) (1999) Exocyclic DNA Adducts in
Carcinogenesis and Mutagenesis. IARC Scienti®c Publications no. 150.
IARC, Lyon, France.
7. Hang,B., Singer,B., Margison,G.P. and Elder,R.H. (1997) Targeted
deletion of alkylpurine-DNA-N-glycosylase in mice eliminates repair of
1,N6-ethenoadenine and hypoxanthine but not of 3,N4-ethenocytosine or
8-oxoguanine. Proc. Natl Acad. Sci. USA, 94, 12869±12874.
8. Saparbaev,M., Kleibl,K. and Laval,J. (1995) Escherichia coli,
Saccharomyces cerevisiae, rat and human 3-methyladenine DNA
glycosylases repair 1,N6-ethenoadenine when present in DNA.
Nucleic Acids Res., 23, 3750±3755.
9. Dosanjh,M.K., Roy,R., Mitra,S. and Singer,B. (1994) 1,N6ethenoadenine is preferred over 3-methyladenine as substrate by a cloned
human N-methylpurine DNA glycosylase (3-methyladenine DNA
glycosylase). Biochemistry, 33, 1624±1628.
10. Singer,B., Antoccia,A., Basu,A.K., Dosanjh,M.K., Fraenkel-Conrat,H.,
Gallagher,P.E., Kusmierek,J.T., Qiu,Z.H. and Rydberg,B. (1992) Both
puri®ed human 1,N6-ethenoadenine-binding protein and puri®ed human
Nucleic Acids Research, 2002, Vol. 30 No. 17
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
3-methyladenine-DNA glycosylase act on 1,N6-ethenoadenine and
3-methyladenine. Proc. Natl Acad. Sci. USA, 89, 9386±9390.
Lau,A.Y., Wyatt,M.D., Glassner,B.J., Samson,L.D. and Ellenberger,T.
(2000) Molecular basis for discriminating between normal and damaged
bases by the human alkyladenine glycosylase, AAG. Proc. Natl Acad.
Sci. USA, 97, 13573±13578.
Hang,B., Downing,G., Guliaev,A. and Singer,B. (2002) Novel activity of
Escherichia coli mismatch uracil-DNA glycosylase (Mug) excising
8-(hydroxymethyl)-3,N4-ethenocytosine, a potential product resulting
from glycidaldehyde reaction. Biochemistry, 41, 2158±2165.
Maruenda,H., Chenna,A., Liem,L.K. and Singer,B. (1998) Synthesis of
1,N6-ethano-2¢-deoxyadenosine, a metabolic product of 1,3-bis(2chloroethyl)nitrosourea, and its incorporation into oligomeric DNA.
J. Org. Chem., 63, 4385±4389.
Hang,B., Sagi,J. and Singer,B. (1998) Correlation between sequencedependent glycosylase repair and the thermal stability of oligonucleotide
duplexes containing 1,N6-ethenoadenine. J. Biol. Chem., 273,
33406±33413.
Rydberg,B., Dosanjh,M.K. and Singer,B. (1991) Human cells contain
protein speci®cally binding to a single 1,N6-ethenoadenine in a DNA
fragment. Proc. Natl Acad. Sci. USA, 88, 6839±6842.
O'Connor,T.R. (1993) Puri®cation and characterization of human
3-methyladenine-DNA glycosylase. Nucleic Acids Res., 21, 5561±5569.
Guliaev,A.B., Sagi,J. and Singer,B. (2000) Sequence-dependent
conformational perturbation in DNA duplexes containing an eA´T
mismatch using molecular dynamics simulation. Carcinogenesis, 21,
1727±1736.
Case,D.A., Pearlman,D.A., Caldwell,J.W., Cheatham,T.E., Ross,W.S.,
Simmerling,C.L., Darden,T.A., Merz,K.M., Stanton,R.V., Cheng,A.L.,
Vincent,J.J., Crowley,M., Tsui,V., Radmer,R.J., Duan,J., Pitera,I.,
Massova,G.L., Seibel,G.L., Singh,U.C., Weiner,P.K. and Kolman,P.A.
(1999) AMBER 6. University of California, San Francisco, CA.
Humphrey,W., Dalke,A. and Schulten,K. (1996) VMDÐvisual
molecular dynamics. J. Mol. Graphics, 14, 33±38.
Lavery,R. and Sklenar,H. (1996) CURVES5.1. Helical Analysis of
Irregular Nucleic Acids. Laboratory for Theoretical Biochemistry,
CNRS, Paris.
Habraken,Y., Carter,C.A., Sekiguchi,M. and Ludlum,D.B. (1991)
Release of N2,3-ethanoguanine from haloethylnitrosourea-treated DNA
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
3787
by Escherichia coli 3-methyladenine DNA glycosylase II.
Carcinogenesis, 12, 1971±1974.
Santos,C.D., Kouchakdjian,M., Yarema,K., Basu,A., Essigmann,J. and
Patel,D.J. (1991) NMR studies of the exocyclic 1,N6ethenodeoxyadenosine adduct (edA) opposite deoxyguanosine in a DNA
duplexÐedA(Syn)-dG(Anti) pairing at the lesion site. Biochemistry, 30,
1828±1835.
Kouchakdjian,M., Eisenberg,M., Yarema,K., Basu,A., Essigmann,J. and
Patel,D.J. (1991) NMR studies of the exocyclic 1,N6ethenodeoxyadenosine adduct (edA) opposite thymidine in a DNA
duplexÐnonplanar alignment of edA(Anti) and dT(Anti) at the lesion
site. Biochemistry, 30, 1820±1828.
Parikh,S.S., Mol,C.D. and Tainer,J.A. (1997) Base excision repair
enzyme family portrait: integrating the structure and chemistry of an
entire DNA repair pathway. Structure, 5, 1543±1550.
Mol,C.D., Parikh,S.S., Putnam,C.D., Lo,T.P. and Tainer,J.A. (1999)
DNA repair mechanisms for the recognition and removal of damaged
DNA bases. Annu. Rev. Biophys. Biomol. Struct., 28, 101±128.
Singer,B. and Hang,B. (1997) What structural features determine repair
enzyme speci®city and mechanism in chemically modi®ed DNA? Chem.
Res. Toxicol., 10, 713±732.
Cullinan,D., Johnson,F. and de los Santos,C. (2000) Solution structure of
an 11-mer duplex containing the 3,N4-ethenocytosine adduct opposite
2¢-deoxycytidine: implications for the recognition of exocyclic lesions by
DNA glycosylases. J. Mol. Biol., 296, 851±861.
Zou,Y., Luo,C. and Geacintov,N.E. (2001) Hierarchy of DNA damage
recognition in Escherichia coli nucleotide excision repair. Biochemistry,
40, 2923±2931.
Lau,A.Y., Scharer,O.D., Samson,L., Verdine,G.L. and Ellenberger,E.
(1998) Crystal structure of a human alkylbase-DNA repair enzyme
complexed to DNA: mechanisms for nucleotide ¯ipping and base
excision. Cell, 95, 249±258.
Yamagata,Y., Kato,M., Odawara,K., Tokuno,Y., Nakashima,Y.,
Matsushima,N., Yasumura,K., Tomita,K., Ihara,K., Fujii,Y.,
Nakabeppu,Y., Sekiguchi,M. and Fujii,S. (1996) Three-dimensional
structure of a DNA repair enzyme, 3-methyladenine DNA glycosylase II,
from Escherichia coli. Cell, 86, 311±319.
Labahn,J., Scharer,O.D., Long,A., Ezaznikpay,K., Verdine,G.L. and
Ellenberger,T.E. (1996) Structural basis for the excision repair of
alkylation-damaged DNA. Cell, 86, 321±329.